Web Application (LAMP*) Security Attack and Defense for System Administrators. *LAMP (linux, apache, mysql, php/perl/python) application security.

Transcription

1 Web Application (LAMP*) Security Attack and Defense for System Administrators *LAMP (linux, apache, mysql, php/perl/python) application security.

2 A little intro Robert Rowley Security/Abuse

3 Backups! They are the first priority. They cover your ass. Asset for quick code comparison. Have multiple backups (off site preferably) Store on read-only media (WORM; write once read many) Don't assume anyone else is already handled it. But how should I back up my data?

4 Beyond the basics. Try a version control option rsync/tar work fine, but CVS, SVN, GIT, etc... can do more! Use them, master them, love them. They not only allow you to identify and quickly correct bugs, flaws, etc... caused by bad code (and help you identify bad coders). They can be used to mitigate attacks, allowing quick distribution of clean code. Not only for code! Use SVN for configuration files, etc.. Having a code repository does not mean you do not need to keep backups! Backup your repository too (now use rsync/tar etc..)!

5 Scenario: Assessing and securing a (LAMP) web application as soon as possible when changing code is not an option. Attack Defense (Hired Pen test) (Sys Admin) Cross Site Scripting (XSS) Application level (mod_sec) SQL injection Network level (snort) Insecure Code File tampering detection

6 Attack: Cross Site Scripting (XSS) Type 1: Reflected attacks The payload exists in the URL and the server side code re-prints the malicious content. (echo $_GET['varname'];) Type 2: Stored attacks The XSS payload is stored on the server (in MySQL, files etc..) and every subsequent request to the same page displays the injected payload. Type 3: DOM attacks Adjusting DOM attributes on the client's browser directly. (When designers go bad! javascript adjust data directly, nothing is actually handled in the server side code itself) Affect: Anyone who visits the page/link is initiating an attack on third party web servers Mis-information, changing/reposting information on a credible website Leaked information (javascript can access cookie information, HTTP variables etc..)

7 Leads to: Leads to: This really is just fun stuff, know enough javascript and you can cause major havoc or do a 0 height iframe for click jacking

8 FUN

9 XSS recap While limited in it's abilities, that is it's strength. Most developers do not consider XSS a security risk leaving a plethora of vulnerable sites. Familiarize yourself with the slight differences between persistent, reflected and DOM XSS attacks Study javascript and be multiple browser compliant.

10 Defense: mod_security Apache module to quickly mitigate and prevent identified attacks.

11 You will need to first install mod_security and then include it in your apache configuration. Instructions vary, but modsecurity.org has what you will need. (example httpd.conf changes after the.so has been compiled) LoadModule security2_module modules/mod_security2.so SecRuleEngine On Include conf/modsecurity.conf... Now all you will need are rules! There are extensive rule sets available at gotroot.com and owasp.org but... it is always best to roll your own. Here is a list of variables you will have available to scan/review: ARGS ARGS_COMBINED_SIZE ARGS_NAMES REQBODY_PROCESSOR REQBODY_ERROR REQBODY_ERROR_MSG XML WEBSERVER_ERROR_LOG FILES FILES_TMPNAMES FILES_NAMES FILES_SIZES FILES_COMBINED_SIZE ENV REMOTE_HOST REMOTE_ADDR REMOTE_PORT REMOTE_USER PATH_INFO QUERY_STRING AUTH_TYPE SERVER_NAME SERVER_ADDR SERVER_PORT TIME_YEAR TIME_EPOCH TIME_MON TIME_DAY TIME_HOUR TIME_MIN TIME_SEC TIME_WDAY TIME REQUEST_URI REQUEST_URI_RAW REQUEST_LINE REQUEST_METHOD REQUEST_PROTOCOL REQUEST_FILENAME REQUEST_BASENAME SCRIPT_FILENAME SCRIPT_BASENAME SCRIPT_UID SCRIPT_GID SCRIPT_USERNAME SCRIPT_GROUPNAME SCRIPT_MODE ENV REQUEST_HEADERS REQUEST_HEADERS_NAMES REQUEST_COOKIES REQUEST_COOKIES_NAMES REQUEST_BODY RESPONSE_LINE RESPONSE_STATUS RESPONSE_PROTOCOL RESPONSE_HEADERS RESPONSE_HEADERS_NAMES RESPONSE_BODY RULE SESSION WEBAPPID SESSIONID USERID

12 Here is a quick rule to stop those simple script/iframe injections via the WP comment form: (contents of conf/modsecurity.conf) SecRule REQUEST_FILENAME wp-comments-post.php chain,deny SecRule ARGS:comment "(<script <iframe)" Check if the requested filename is wp-comments-post.php and chain (continue checking) with the next rule... Check if the argument (POST or GET) named comment has the string <script or <iframe. If both are true then take the specified action (deny) Over-zealous version: SecRule ARGS "(<script <iframe)" deny Any argument (POST/GET variable) that matches the pattern <script or <iframe will be denied.

13 Remember from earlier? Leads to: Leads to:

14 mod_security recap Pro Prevents attacks from succeeding Application level access (no worries about SSL needing to be intercepted/decrypted) Con Only as powerful as the rules used Bad rules will create false positives. Very flexible rule sets Extendibility with scripting (something for another time) Detectable by attackers, who can adjust their attacks (something for another time)

15 Attack: SQL injection Do (almost) anything you want to the database!

16 SQL injection with photoracer plugin.

17 Simplest form, just add a union statement added to one of the GET/POST variables that gets appended to the select query

18 Here is what went wrong in the code: 16 $imgid = $_REQUEST['id']; 17 $q1 = "SELECT raceid, wpuid, imgid, imgpath, imgname, imgcomment, sumvotes, imgcountview, tinsert FROM ". 18 $wpdb->prefix."photoracer WHERE imgid=$imgid"; $out = $wpdb->get_row($q1); "da :".get_author_name($out->wpuid)."<br/>". $out->imgcomment."<br/>". Expected SQL statement: SELECT raceid, wpuid, imgid, imgpath, imgname, imgcomment, sumvotes, imgcountview, tinsert FROM wp_photoracer WHERE imgid=10 Injected SQL statement: SELECT raceid, wpuid, imgid, imgpath, imgname, imgcomment, sumvotes, imgcountview, tinsert FROM wp_photoracer WHERE imgid=-1 union select 0,1,2,3,4,user_login,6,7,8 from wp_users--

19 But wait... there is more. Let's get some password hashes!

20 SQL injection recap Requires an understanding of SQL statements Very noisy if the database structure is not known Time consuming (if not automated) Can do more than SELECT based on user's privileges! (DROP, ALTER, CREATE, GRANT statements are all possible.)

21 DEFENSE: IDS/IPS (snort and more) Excellent method to monitor network traffic and retrieve information on attacks. Don't forget your whitelist!

22 Snort is a widely used IDS (intrusion detection system) Lightweight Large user base for example rulesets and help Snort can detect more than just web application attacks (unlike mod_security.) Setup is easy: download, compile, configure, monitor and update! By itself snort will only log the attacks for review, plugins like SnortSam/Guardian.pl will turn snort into a powerful IPS (intrusion prevention system)

23 Example code to detect the SQL injection attack: alert tcp $EXTERNAL_NET any -> $HTTP_SERVERS $HTTP_PORTS ( msg:"sql-injection photoracer Sql Injection attempt"; flow:to_server,established; uricontent:"viewimg.php"; nocase; uricontent:"id="; uricontent:"union"; nocase; uricontent:"select"; nocase; classtype:web-application-attack; sid: ; rev:2;) (detects /viewimg.php... id=...union...select... What the attack looks like in the logs: [**] [1: :2] SQL-INJECTION photoracer Sql Injection attempt [**] [Classification: Web Application Attack] [Priority: 1] 08/08-23:26: : > :80 TCP TTL:55 TOS:0x0 ID:63052 IpLen:20 DgmLen:444 DF ***AP*** Seq: 0x4A59450C Ack: 0x9CB41D1A Win: 0x16D0 TcpLen: 20

24 SNORTSAM/Guardian.pl These scripts watch the snort log files, upon evidence of an attack they will use iptables/ipfwadm etc... to firewall the attacker's IP. The most important thing to remember is: whitelist your system's IP if you plan to test new rules. (this prevents you from locking yourself out of the server) Example: Using a script like snortsam/guardian.pl the attacker's IP is blocked. # iptables -n -L Chain INPUT (policy ACCEPT) target prot opt source DROP all destination /0...

25 Snort recap Pros Cons Lightweight Only as good as your rules Runs independently False positives Powerful as an IDS or IPS Can become very daunting to customize Requires third party scripts to be an IPS Network level can see more than application level(mod_security) Large user base, lots of help available As an IPS will stop the attacker in their tracks (at least their IP) False positives when running a IPS may leave your IP blocked, don't forget to whitelist yourself!

26 Attack: Code Vulnerability Code can be attacked! The #1 fault is, trusting user input. RISK: Attackers get to do whatever they want.

27 Attack: Badly Coded Upload form! Allowing people to upload files willy nilly will get you compromised quickly. RISK: Allowing users to upload executable files (.php.pl.cgi etc..) or even.html files is giving away the keys to the castle.

28 // Where the file is going to be placed $target_path = "images/"; /* Add the original filename to our target path. Result is "uploads/filename.extension" */ $target_path = $target_path. basename( $_FILES['uploadedfile']['name']); if(move_uploaded_file($_files['uploadedfile']['tmp_name'], $target_path)) { echo "The file ". basename( $_FILES['uploadedfile']['name']). " has been uploaded"; } else{ echo "There was an error uploading the file, please try again!"; } The flaw? No checking the file's extension or checking if it is really an image image, movie, etc... Find any php shell ( and upload for FUN!

29 The upload page Example backdoor

30 DEFENSE: File monitoring inotify also: (tripwire, fschange, kqueue utilities etc..)

31 Inotify an you... Kernel module (released in ) Monitors file system changes Improvement from dnotify Inotify-tools, incron, iwatch, pynotify

32 inotify-tools libinotifytools inotifywait inotifywatch

33 inotify/incron quick and easy 1. Verify your kernel has inotify enabled, install incrond 2. Setup incrontab to monitor fies/directores i. (directory) (mask) (command) 3. Start/Restart incrond Example incrontab: /home/user/website IN_CREATE mail -s 'File created!'

34 More about masks IN_ACCESS IN_ATTRIB IN_CLOSE_WRITE IN_CLOSE_NOWRITE IN_CREATE IN_DELETE IN_DELETE_SELF IN_MODIFY IN_MOVE_SELF IN_MOVED_FROM IN_MOVED_TO IN_OPEN File was accessed (read) Metadata changed (permissions, timestamps, extended attributes, etc.) File opened for writing was closed File not opened for writing was closed File/directory created in watched directory File/directory deleted from watched directory Watched file/directory was itself deleted File was modified Watched file/directory was itself moved File moved out of watched directory File moved into watched directory File was opened

35 Passing Variables The command may contain these wildcards: $$ - a dollar sign $@ - the watched filesystem path (see above) $# - the event-related file name $% - the event flags (textually) $& - the event flags (numerically) Example incron entry: /directory/to/watch IN_MODIFY /path/to/script $@/$# This will execute the script and pass it the file name which triggered the rule

36 A diff erent solution... Compare your live data to a backup! A utility you are probably familiar with diff can do this for you. Just run diff -r (live directory) (backup directory) You will receive a report of file differences.

37 DEFENSE: Forensics It's Logs, it's Logs It's better than bad, it's good. Logs, logs, logs!

38 Useful logs: Apache (website) Auth.log, FTP.log Syslog / messages Specific IDS logs Tools to know: Grep Awk Perl/shell scripting

39 Coorealation of timestamps is key (most of the time) Use NTP or a central logging server. Previous attack (detected by tripwire) Modified object name: /home/user/website/upload/images/c100.php Property: Expected Observed * Inode Number * Modify Time Sun 15 Aug :24:38 PM PDT Sun 15 Aug :26:49 PM PDT * Change Time Sun 15 Aug :24:38 PM PDT Sun 15 Aug :26:49 PM PDT Data found in the logs: debian:~/apache_logs$ grep 18:24: wp-access.log [15/Aug/2010:18:24: ] "POST /upload/uploader.php HTTP/1.1" 200 Confirm the file exists on the server: debian:~$ ls -la /home/user/website/upload/images/c100.php -rw-r--r-- 1 www-data www-data :26 /home/user/website/upload/images/c100.php

40 Wrap up Writing secure code will prevent the need for all of this. Until that day comes, enjoy what you have learned and apply it. Attacks are mostly automated; happening daily, hourly, right now...

41 From here on out... You're on your own to continue researching the subject matters covered briefly in this talk. I have a list of URLs you are free to copy, as well as ideas for more talks if anyone wants to throw their hat in the ring. Writing secure code! Penetration testing! Why your firewall is racist. SELECT MBA!= DBA.

42 Further reading! General Security (Excellent resource for many security topics) (Pay attention to the latest security concerns) (Exploit/tools resource) Snort (IDS) (Turns snort into an IPS) mod_security (Web application level firewall) (mod_security rule list) inotify (file change detection) (incron etc..) (inotify toolset) Cross Site scripting: (Cross site scripting attacks archive) Backdoor web shells: (list of backdoor shells) SQL Injection (plenty of information available)